Assessing the state of substitution models describing noncoding RNA evolution.

Area of Science:

• Phylogenetics and evolutionary biology
• Bioinformatics and computational biology
• Molecular evolution

Background:

• Phylogenetic inference relies on accurate substitution models, with RNA evolution models varying in state space (7-state vs. 16-state).
• Current model selection methods struggle to compare models with different state spaces due to likelihood conditioning.
• RNA molecules are crucial for phylogenetic studies due to their ubiquity and slow evolution.

Purpose of the Study:

• To introduce statistical state-space projection methods for comparing RNA and nucleotide phylogenetic models.
• To enable direct likelihood comparisons across models with different state spaces.
• To facilitate robust model selection for RNA sequence evolution.

Main Methods:

• Development of statistical state-space projection methods.
• Application of methods to 287 RNA families from genomic alignments.
• Model selection performed using the open-source PHASE 3.0 software.

Main Results:

• RNA models were selected over nucleotide models in 281 out of 287 RNA families.
• 7-state RNA models favored for conserved families with short stems.
• 16-state RNA models favored for divergent families with long stems.
• RNA function and GC-content had minimal impact on model selection.

Conclusions:

• Statistical state-space projection effectively enables cross-model selection for RNA phylogenetic inference.
• The choice between 7-state and 16-state RNA models depends on sequence conservation and structural features.
• The developed methods and software provide a valuable tool for studying RNA evolution.